This paper illustrates the design and implementation of a prototype ASP solver that is capable of exploiting the parallelism offered by general purpose graphical processing units (GPGPUs). The solver is based on a basic conflict-driven search algorithm. The core of the solving process develops on the CPU, while most of the activities, such as literal selection, unit propagation, and conflict-analysis, are delegated to the GPU. Moreover, a deep non-deterministic search, involving a very large number of threads, is also delegated to the GPU. The initial results confirm the feasibility of the approach and the potential offered by GPUs in the context of ASP computations.
CUD@ASP: Experimenting with GPGPUs in ASP solving
FORMISANO, Andrea;
2013
Abstract
This paper illustrates the design and implementation of a prototype ASP solver that is capable of exploiting the parallelism offered by general purpose graphical processing units (GPGPUs). The solver is based on a basic conflict-driven search algorithm. The core of the solving process develops on the CPU, while most of the activities, such as literal selection, unit propagation, and conflict-analysis, are delegated to the GPU. Moreover, a deep non-deterministic search, involving a very large number of threads, is also delegated to the GPU. The initial results confirm the feasibility of the approach and the potential offered by GPUs in the context of ASP computations.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
